File size: 1,911 Bytes
cac0037
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
language: en-US
early_access: false
reviews:
  profile: assertive
  request_changes_workflow: true
  high_level_summary: true
  high_level_summary_instructions: |
    Summarize this PR as work on NEXUS Visual Weaver, a Hugging Face Build Small
    Hackathon Gradio command-center for governed visual creation. Lead with
    user-facing dashboard/workflow impact, then model-governance impact, then
    security/export-gate impact, then verification evidence.

    Preserve these project anchors in summaries when relevant:
    - FLUX.2 is the pinned image-generation lane.
    - NVIDIA LocateAnything-3B is the pinned grounding lane.
    - ST3GG is the always-on security/export gate.
    - Adult Mode must remain opt-in and must not disable provenance, consent,
      age, export, dataset-partition, or ST3GG gates.
    - ModelRelay/GMR helper rotation is quota-aware and only applies to helper
      lanes, not pinned core lanes.

    Avoid marketing filler, poems, exaggerated claims, or claims that provider
    calls actually ran unless tests or logs prove it. Distinguish implemented
    runtime behavior from dry-run preview surfaces and planning scaffolds.
  poem: false
  review_status: true
  collapse_walkthrough: false
  path_filters:
    - "!outputs/**"
    - "!**/__pycache__/**"
    - "!pytest-cache-files-*/**"
  path_instructions:
    - path: "src/nexus_visual_weaver/**"
      instructions: "Focus on Gradio runtime correctness, defensive security gates, model-lane governance, parameter-budget logic, and test coverage for fallback behavior."
    - path: "app.py"
      instructions: "Check that UI callbacks update the intended regions, never leak secrets, and keep Adult Mode security gates active."
    - path: "tests/**"
      instructions: "Prefer focused regression tests for workflow routing, ST3GG evidence, catalog filtering, and ModelRelay quota behavior."
chat:
  auto_reply: true